drm/amdgpu: add option params to enforce process isolation between graphics and compute

enforce process isolation between graphics and compute via using the same reserved vmid.

v2: remove params "struct amdgpu_vm *vm" from
    amdgpu_vmid_alloc_reserved and amdgpu_vmid_free_reserved.

Signed-off-by: Chong Li <chongli2@amd.com>
Reviewed-by: Christian Koenig <Christian.Koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Author: Chong Li

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -214,6 +214,7 @@ extern int amdgpu_force_asic_type;
 extern int amdgpu_smartshift_bias;
 extern int amdgpu_use_xgmi_p2p;
 extern int amdgpu_mtype_local;
+extern bool enforce_isolation;
 #ifdef CONFIG_HSA_AMD
 extern int sched_policy;
 extern bool debug_evictions;

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -153,7 +153,7 @@ uint amdgpu_pg_mask = 0xffffffff;
 uint amdgpu_sdma_phase_quantum = 32;
 char *amdgpu_disable_cu;
 char *amdgpu_virtual_display;
-
+bool enforce_isolation;
 /*
  * OverDrive(bit 14) disabled by default
  * GFX DCS(bit 19) disabled by default
@@ -973,6 +973,14 @@ MODULE_PARM_DESC(
 						4 = AMDGPU_CPX_PARTITION_MODE)");
 module_param_named(user_partt_mode, amdgpu_user_partt_mode, uint, 0444);
 
+
+/**
+ * DOC: enforce_isolation (bool)
+ * enforce process isolation between graphics and compute via using the same reserved vmid.
+ */
+module_param(enforce_isolation, bool, 0444);
+MODULE_PARM_DESC(enforce_isolation, "enforce process isolation between graphics and compute . enforce_isolation = on");
+
 /* These devices are not supported by amdgpu.
  * They are supported by the mach64, r128, radeon drivers
  */

drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c

@@ -409,7 +409,7 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 	if (r || !idle)
 		goto error;
 
-	if (vm->reserved_vmid[vmhub]) {
+	if (vm->reserved_vmid[vmhub] || (enforce_isolation && (vmhub == AMDGPU_GFXHUB(0)))) {
 		r = amdgpu_vmid_grab_reserved(vm, ring, job, &id, fence);
 		if (r || !id)
 			goto error;
@@ -460,14 +460,11 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 }
 
 int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
 			       unsigned vmhub)
 {
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 
 	mutex_lock(&id_mgr->lock);
-	if (vm->reserved_vmid[vmhub])
-		goto unlock;
 
 	++id_mgr->reserved_use_count;
 	if (!id_mgr->reserved) {
@@ -479,27 +476,23 @@ int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
 		list_del_init(&id->list);
 		id_mgr->reserved = id;
 	}
-	vm->reserved_vmid[vmhub] = true;
 
-unlock:
 	mutex_unlock(&id_mgr->lock);
 	return 0;
 }
 
 void amdgpu_vmid_free_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
 			       unsigned vmhub)
 {
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 
 	mutex_lock(&id_mgr->lock);
-	if (vm->reserved_vmid[vmhub] &&
-	    !--id_mgr->reserved_use_count) {
+	if (!--id_mgr->reserved_use_count) {
 		/* give the reserved ID back to normal round robin */
 		list_add(&id_mgr->reserved->list, &id_mgr->ids_lru);
 		id_mgr->reserved = NULL;
 	}
-	vm->reserved_vmid[vmhub] = false;
+
 	mutex_unlock(&id_mgr->lock);
 }
 
@@ -578,6 +571,10 @@ void amdgpu_vmid_mgr_init(struct amdgpu_device *adev)
 			list_add_tail(&id_mgr->ids[j].list, &id_mgr->ids_lru);
 		}
 	}
+	/* alloc a default reserved vmid to enforce isolation */
+	if (enforce_isolation)
+		amdgpu_vmid_alloc_reserved(adev, AMDGPU_GFXHUB(0));
+
 }
 
 /**

drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h

@@ -79,11 +79,9 @@ void amdgpu_pasid_free_delayed(struct dma_resv *resv,
 bool amdgpu_vmid_had_gpu_reset(struct amdgpu_device *adev,
 			       struct amdgpu_vmid *id);
 int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
-			       unsigned vmhub);
+				unsigned vmhub);
 void amdgpu_vmid_free_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
-			       unsigned vmhub);
+				unsigned vmhub);
 int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 		     struct amdgpu_job *job, struct dma_fence **fence);
 void amdgpu_vmid_reset(struct amdgpu_device *adev, unsigned vmhub,

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

@@ -2284,8 +2284,14 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	}
 
 	dma_fence_put(vm->last_update);
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
-		amdgpu_vmid_free_reserved(adev, vm, i);
+
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) {
+		if (vm->reserved_vmid[i]) {
+			amdgpu_vmid_free_reserved(adev, i);
+			vm->reserved_vmid[i] = false;
+		}
+	}
+
 }
 
 /**
@@ -2368,7 +2374,6 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	union drm_amdgpu_vm *args = data;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_fpriv *fpriv = filp->driver_priv;
-	int r;
 
 	/* No valid flags defined yet */
 	if (args->in.flags)
@@ -2377,13 +2382,17 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	switch (args->in.op) {
 	case AMDGPU_VM_OP_RESERVE_VMID:
 		/* We only have requirement to reserve vmid from gfxhub */
-		r = amdgpu_vmid_alloc_reserved(adev, &fpriv->vm,
-					       AMDGPU_GFXHUB(0));
-		if (r)
-			return r;
+		if (!fpriv->vm.reserved_vmid[AMDGPU_GFXHUB(0)]) {
+			amdgpu_vmid_alloc_reserved(adev, AMDGPU_GFXHUB(0));
+			fpriv->vm.reserved_vmid[AMDGPU_GFXHUB(0)] = true;
+		}
+
 		break;
 	case AMDGPU_VM_OP_UNRESERVE_VMID:
-		amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB(0));
+		if (fpriv->vm.reserved_vmid[AMDGPU_GFXHUB(0)]) {
+			amdgpu_vmid_free_reserved(adev, AMDGPU_GFXHUB(0));
+			fpriv->vm.reserved_vmid[AMDGPU_GFXHUB(0)] = false;
+		}
 		break;
 	default:
 		return -EINVAL;